1/*
2 * Copyright (C) 2001-2002 Sistina Software (UK) Limited.
3 * Copyright (C) 2006-2008 Red Hat GmbH
4 *
5 * This file is released under the GPL.
6 */
7
8#include "dm-exception-store.h"
9
10#include <linux/ctype.h>
11#include <linux/mm.h>
12#include <linux/pagemap.h>
13#include <linux/vmalloc.h>
14#include <linux/export.h>
15#include <linux/slab.h>
16#include <linux/dm-io.h>
17#include <linux/dm-bufio.h>
18
19#define DM_MSG_PREFIX "persistent snapshot"
20#define DM_CHUNK_SIZE_DEFAULT_SECTORS 32U	/* 16KB */
21
22#define DM_PREFETCH_CHUNKS		12
23
24/*-----------------------------------------------------------------
25 * Persistent snapshots, by persistent we mean that the snapshot
26 * will survive a reboot.
27 *---------------------------------------------------------------*/
28
29/*
30 * We need to store a record of which parts of the origin have
31 * been copied to the snapshot device.  The snapshot code
32 * requires that we copy exception chunks to chunk aligned areas
33 * of the COW store.  It makes sense therefore, to store the
34 * metadata in chunk size blocks.
35 *
36 * There is no backward or forward compatibility implemented,
37 * snapshots with different disk versions than the kernel will
38 * not be usable.  It is expected that "lvcreate" will blank out
39 * the start of a fresh COW device before calling the snapshot
40 * constructor.
41 *
42 * The first chunk of the COW device just contains the header.
43 * After this there is a chunk filled with exception metadata,
44 * followed by as many exception chunks as can fit in the
45 * metadata areas.
46 *
47 * All on disk structures are in little-endian format.  The end
48 * of the exceptions info is indicated by an exception with a
49 * new_chunk of 0, which is invalid since it would point to the
50 * header chunk.
51 */
52
53/*
54 * Magic for persistent snapshots: "SnAp" - Feeble isn't it.
55 */
56#define SNAP_MAGIC 0x70416e53
57
58/*
59 * The on-disk version of the metadata.
60 */
61#define SNAPSHOT_DISK_VERSION 1
62
63#define NUM_SNAPSHOT_HDR_CHUNKS 1
64
65struct disk_header {
66	__le32 magic;
67
68	/*
69	 * Is this snapshot valid.  There is no way of recovering
70	 * an invalid snapshot.
71	 */
72	__le32 valid;
73
74	/*
75	 * Simple, incrementing version. no backward
76	 * compatibility.
77	 */
78	__le32 version;
79
80	/* In sectors */
81	__le32 chunk_size;
82} __packed;
83
84struct disk_exception {
85	__le64 old_chunk;
86	__le64 new_chunk;
87} __packed;
88
89struct core_exception {
90	uint64_t old_chunk;
91	uint64_t new_chunk;
92};
93
94struct commit_callback {
95	void (*callback)(void *, int success);
96	void *context;
97};
98
99/*
100 * The top level structure for a persistent exception store.
101 */
102struct pstore {
103	struct dm_exception_store *store;
104	int version;
105	int valid;
106	uint32_t exceptions_per_area;
107
108	/*
109	 * Now that we have an asynchronous kcopyd there is no
110	 * need for large chunk sizes, so it wont hurt to have a
111	 * whole chunks worth of metadata in memory at once.
112	 */
113	void *area;
114
115	/*
116	 * An area of zeros used to clear the next area.
117	 */
118	void *zero_area;
119
120	/*
121	 * An area used for header. The header can be written
122	 * concurrently with metadata (when invalidating the snapshot),
123	 * so it needs a separate buffer.
124	 */
125	void *header_area;
126
127	/*
128	 * Used to keep track of which metadata area the data in
129	 * 'chunk' refers to.
130	 */
131	chunk_t current_area;
132
133	/*
134	 * The next free chunk for an exception.
135	 *
136	 * When creating exceptions, all the chunks here and above are
137	 * free.  It holds the next chunk to be allocated.  On rare
138	 * occasions (e.g. after a system crash) holes can be left in
139	 * the exception store because chunks can be committed out of
140	 * order.
141	 *
142	 * When merging exceptions, it does not necessarily mean all the
143	 * chunks here and above are free.  It holds the value it would
144	 * have held if all chunks had been committed in order of
145	 * allocation.  Consequently the value may occasionally be
146	 * slightly too low, but since it's only used for 'status' and
147	 * it can never reach its minimum value too early this doesn't
148	 * matter.
149	 */
150
151	chunk_t next_free;
152
153	/*
154	 * The index of next free exception in the current
155	 * metadata area.
156	 */
157	uint32_t current_committed;
158
159	atomic_t pending_count;
160	uint32_t callback_count;
161	struct commit_callback *callbacks;
162	struct dm_io_client *io_client;
163
164	struct workqueue_struct *metadata_wq;
165};
166
167static int alloc_area(struct pstore *ps)
168{
169	int r = -ENOMEM;
170	size_t len;
171
172	len = ps->store->chunk_size << SECTOR_SHIFT;
173
174	/*
175	 * Allocate the chunk_size block of memory that will hold
176	 * a single metadata area.
177	 */
178	ps->area = vmalloc(len);
179	if (!ps->area)
180		goto err_area;
181
182	ps->zero_area = vzalloc(len);
183	if (!ps->zero_area)
184		goto err_zero_area;
185
186	ps->header_area = vmalloc(len);
187	if (!ps->header_area)
188		goto err_header_area;
189
190	return 0;
191
192err_header_area:
193	vfree(ps->zero_area);
194
195err_zero_area:
196	vfree(ps->area);
197
198err_area:
199	return r;
200}
201
202static void free_area(struct pstore *ps)
203{
204	vfree(ps->area);
205	ps->area = NULL;
206	vfree(ps->zero_area);
207	ps->zero_area = NULL;
208	vfree(ps->header_area);
209	ps->header_area = NULL;
210}
211
212struct mdata_req {
213	struct dm_io_region *where;
214	struct dm_io_request *io_req;
215	struct work_struct work;
216	int result;
217};
218
219static void do_metadata(struct work_struct *work)
220{
221	struct mdata_req *req = container_of(work, struct mdata_req, work);
222
223	req->result = dm_io(req->io_req, 1, req->where, NULL);
224}
225
226/*
227 * Read or write a chunk aligned and sized block of data from a device.
228 */
229static int chunk_io(struct pstore *ps, void *area, chunk_t chunk, int op,
230		    int op_flags, int metadata)
231{
232	struct dm_io_region where = {
233		.bdev = dm_snap_cow(ps->store->snap)->bdev,
234		.sector = ps->store->chunk_size * chunk,
235		.count = ps->store->chunk_size,
236	};
237	struct dm_io_request io_req = {
238		.bi_op = op,
239		.bi_op_flags = op_flags,
240		.mem.type = DM_IO_VMA,
241		.mem.ptr.vma = area,
242		.client = ps->io_client,
243		.notify.fn = NULL,
244	};
245	struct mdata_req req;
246
247	if (!metadata)
248		return dm_io(&io_req, 1, &where, NULL);
249
250	req.where = &where;
251	req.io_req = &io_req;
252
253	/*
254	 * Issue the synchronous I/O from a different thread
255	 * to avoid submit_bio_noacct recursion.
256	 */
257	INIT_WORK_ONSTACK(&req.work, do_metadata);
258	queue_work(ps->metadata_wq, &req.work);
259	flush_workqueue(ps->metadata_wq);
260	destroy_work_on_stack(&req.work);
261
262	return req.result;
263}
264
265/*
266 * Convert a metadata area index to a chunk index.
267 */
268static chunk_t area_location(struct pstore *ps, chunk_t area)
269{
270	return NUM_SNAPSHOT_HDR_CHUNKS + ((ps->exceptions_per_area + 1) * area);
271}
272
273static void skip_metadata(struct pstore *ps)
274{
275	uint32_t stride = ps->exceptions_per_area + 1;
276	chunk_t next_free = ps->next_free;
277	if (sector_div(next_free, stride) == NUM_SNAPSHOT_HDR_CHUNKS)
278		ps->next_free++;
279}
280
281/*
282 * Read or write a metadata area.  Remembering to skip the first
283 * chunk which holds the header.
284 */
285static int area_io(struct pstore *ps, int op, int op_flags)
286{
287	chunk_t chunk = area_location(ps, ps->current_area);
288
289	return chunk_io(ps, ps->area, chunk, op, op_flags, 0);
290}
291
292static void zero_memory_area(struct pstore *ps)
293{
294	memset(ps->area, 0, ps->store->chunk_size << SECTOR_SHIFT);
295}
296
297static int zero_disk_area(struct pstore *ps, chunk_t area)
298{
299	return chunk_io(ps, ps->zero_area, area_location(ps, area),
300			REQ_OP_WRITE, 0, 0);
301}
302
303static int read_header(struct pstore *ps, int *new_snapshot)
304{
305	int r;
306	struct disk_header *dh;
307	unsigned chunk_size;
308	int chunk_size_supplied = 1;
309	char *chunk_err;
310
311	/*
312	 * Use default chunk size (or logical_block_size, if larger)
313	 * if none supplied
314	 */
315	if (!ps->store->chunk_size) {
316		ps->store->chunk_size = max(DM_CHUNK_SIZE_DEFAULT_SECTORS,
317		    bdev_logical_block_size(dm_snap_cow(ps->store->snap)->
318					    bdev) >> 9);
319		ps->store->chunk_mask = ps->store->chunk_size - 1;
320		ps->store->chunk_shift = __ffs(ps->store->chunk_size);
321		chunk_size_supplied = 0;
322	}
323
324	ps->io_client = dm_io_client_create();
325	if (IS_ERR(ps->io_client))
326		return PTR_ERR(ps->io_client);
327
328	r = alloc_area(ps);
329	if (r)
330		return r;
331
332	r = chunk_io(ps, ps->header_area, 0, REQ_OP_READ, 0, 1);
333	if (r)
334		goto bad;
335
336	dh = ps->header_area;
337
338	if (le32_to_cpu(dh->magic) == 0) {
339		*new_snapshot = 1;
340		return 0;
341	}
342
343	if (le32_to_cpu(dh->magic) != SNAP_MAGIC) {
344		DMWARN("Invalid or corrupt snapshot");
345		r = -ENXIO;
346		goto bad;
347	}
348
349	*new_snapshot = 0;
350	ps->valid = le32_to_cpu(dh->valid);
351	ps->version = le32_to_cpu(dh->version);
352	chunk_size = le32_to_cpu(dh->chunk_size);
353
354	if (ps->store->chunk_size == chunk_size)
355		return 0;
356
357	if (chunk_size_supplied)
358		DMWARN("chunk size %u in device metadata overrides "
359		       "table chunk size of %u.",
360		       chunk_size, ps->store->chunk_size);
361
362	/* We had a bogus chunk_size. Fix stuff up. */
363	free_area(ps);
364
365	r = dm_exception_store_set_chunk_size(ps->store, chunk_size,
366					      &chunk_err);
367	if (r) {
368		DMERR("invalid on-disk chunk size %u: %s.",
369		      chunk_size, chunk_err);
370		return r;
371	}
372
373	r = alloc_area(ps);
374	return r;
375
376bad:
377	free_area(ps);
378	return r;
379}
380
381static int write_header(struct pstore *ps)
382{
383	struct disk_header *dh;
384
385	memset(ps->header_area, 0, ps->store->chunk_size << SECTOR_SHIFT);
386
387	dh = ps->header_area;
388	dh->magic = cpu_to_le32(SNAP_MAGIC);
389	dh->valid = cpu_to_le32(ps->valid);
390	dh->version = cpu_to_le32(ps->version);
391	dh->chunk_size = cpu_to_le32(ps->store->chunk_size);
392
393	return chunk_io(ps, ps->header_area, 0, REQ_OP_WRITE, 0, 1);
394}
395
396/*
397 * Access functions for the disk exceptions, these do the endian conversions.
398 */
399static struct disk_exception *get_exception(struct pstore *ps, void *ps_area,
400					    uint32_t index)
401{
402	BUG_ON(index >= ps->exceptions_per_area);
403
404	return ((struct disk_exception *) ps_area) + index;
405}
406
407static void read_exception(struct pstore *ps, void *ps_area,
408			   uint32_t index, struct core_exception *result)
409{
410	struct disk_exception *de = get_exception(ps, ps_area, index);
411
412	/* copy it */
413	result->old_chunk = le64_to_cpu(de->old_chunk);
414	result->new_chunk = le64_to_cpu(de->new_chunk);
415}
416
417static void write_exception(struct pstore *ps,
418			    uint32_t index, struct core_exception *e)
419{
420	struct disk_exception *de = get_exception(ps, ps->area, index);
421
422	/* copy it */
423	de->old_chunk = cpu_to_le64(e->old_chunk);
424	de->new_chunk = cpu_to_le64(e->new_chunk);
425}
426
427static void clear_exception(struct pstore *ps, uint32_t index)
428{
429	struct disk_exception *de = get_exception(ps, ps->area, index);
430
431	/* clear it */
432	de->old_chunk = 0;
433	de->new_chunk = 0;
434}
435
436/*
437 * Registers the exceptions that are present in the current area.
438 * 'full' is filled in to indicate if the area has been
439 * filled.
440 */
441static int insert_exceptions(struct pstore *ps, void *ps_area,
442			     int (*callback)(void *callback_context,
443					     chunk_t old, chunk_t new),
444			     void *callback_context,
445			     int *full)
446{
447	int r;
448	unsigned int i;
449	struct core_exception e;
450
451	/* presume the area is full */
452	*full = 1;
453
454	for (i = 0; i < ps->exceptions_per_area; i++) {
455		read_exception(ps, ps_area, i, &e);
456
457		/*
458		 * If the new_chunk is pointing at the start of
459		 * the COW device, where the first metadata area
460		 * is we know that we've hit the end of the
461		 * exceptions.  Therefore the area is not full.
462		 */
463		if (e.new_chunk == 0LL) {
464			ps->current_committed = i;
465			*full = 0;
466			break;
467		}
468
469		/*
470		 * Keep track of the start of the free chunks.
471		 */
472		if (ps->next_free <= e.new_chunk)
473			ps->next_free = e.new_chunk + 1;
474
475		/*
476		 * Otherwise we add the exception to the snapshot.
477		 */
478		r = callback(callback_context, e.old_chunk, e.new_chunk);
479		if (r)
480			return r;
481	}
482
483	return 0;
484}
485
486static int read_exceptions(struct pstore *ps,
487			   int (*callback)(void *callback_context, chunk_t old,
488					   chunk_t new),
489			   void *callback_context)
490{
491	int r, full = 1;
492	struct dm_bufio_client *client;
493	chunk_t prefetch_area = 0;
494
495	client = dm_bufio_client_create(dm_snap_cow(ps->store->snap)->bdev,
496					ps->store->chunk_size << SECTOR_SHIFT,
497					1, 0, NULL, NULL);
498
499	if (IS_ERR(client))
500		return PTR_ERR(client);
501
502	/*
503	 * Setup for one current buffer + desired readahead buffers.
504	 */
505	dm_bufio_set_minimum_buffers(client, 1 + DM_PREFETCH_CHUNKS);
506
507	/*
508	 * Keeping reading chunks and inserting exceptions until
509	 * we find a partially full area.
510	 */
511	for (ps->current_area = 0; full; ps->current_area++) {
512		struct dm_buffer *bp;
513		void *area;
514		chunk_t chunk;
515
516		if (unlikely(prefetch_area < ps->current_area))
517			prefetch_area = ps->current_area;
518
519		if (DM_PREFETCH_CHUNKS) do {
520			chunk_t pf_chunk = area_location(ps, prefetch_area);
521			if (unlikely(pf_chunk >= dm_bufio_get_device_size(client)))
522				break;
523			dm_bufio_prefetch(client, pf_chunk, 1);
524			prefetch_area++;
525			if (unlikely(!prefetch_area))
526				break;
527		} while (prefetch_area <= ps->current_area + DM_PREFETCH_CHUNKS);
528
529		chunk = area_location(ps, ps->current_area);
530
531		area = dm_bufio_read(client, chunk, &bp);
532		if (IS_ERR(area)) {
533			r = PTR_ERR(area);
534			goto ret_destroy_bufio;
535		}
536
537		r = insert_exceptions(ps, area, callback, callback_context,
538				      &full);
539
540		if (!full)
541			memcpy(ps->area, area, ps->store->chunk_size << SECTOR_SHIFT);
542
543		dm_bufio_release(bp);
544
545		dm_bufio_forget(client, chunk);
546
547		if (unlikely(r))
548			goto ret_destroy_bufio;
549	}
550
551	ps->current_area--;
552
553	skip_metadata(ps);
554
555	r = 0;
556
557ret_destroy_bufio:
558	dm_bufio_client_destroy(client);
559
560	return r;
561}
562
563static struct pstore *get_info(struct dm_exception_store *store)
564{
565	return (struct pstore *) store->context;
566}
567
568static void persistent_usage(struct dm_exception_store *store,
569			     sector_t *total_sectors,
570			     sector_t *sectors_allocated,
571			     sector_t *metadata_sectors)
572{
573	struct pstore *ps = get_info(store);
574
575	*sectors_allocated = ps->next_free * store->chunk_size;
576	*total_sectors = get_dev_size(dm_snap_cow(store->snap)->bdev);
577
578	/*
579	 * First chunk is the fixed header.
580	 * Then there are (ps->current_area + 1) metadata chunks, each one
581	 * separated from the next by ps->exceptions_per_area data chunks.
582	 */
583	*metadata_sectors = (ps->current_area + 1 + NUM_SNAPSHOT_HDR_CHUNKS) *
584			    store->chunk_size;
585}
586
587static void persistent_dtr(struct dm_exception_store *store)
588{
589	struct pstore *ps = get_info(store);
590
591	destroy_workqueue(ps->metadata_wq);
592
593	/* Created in read_header */
594	if (ps->io_client)
595		dm_io_client_destroy(ps->io_client);
596	free_area(ps);
597
598	/* Allocated in persistent_read_metadata */
599	vfree(ps->callbacks);
600
601	kfree(ps);
602}
603
604static int persistent_read_metadata(struct dm_exception_store *store,
605				    int (*callback)(void *callback_context,
606						    chunk_t old, chunk_t new),
607				    void *callback_context)
608{
609	int r, new_snapshot;
610	struct pstore *ps = get_info(store);
611
612	/*
613	 * Read the snapshot header.
614	 */
615	r = read_header(ps, &new_snapshot);
616	if (r)
617		return r;
618
619	/*
620	 * Now we know correct chunk_size, complete the initialisation.
621	 */
622	ps->exceptions_per_area = (ps->store->chunk_size << SECTOR_SHIFT) /
623				  sizeof(struct disk_exception);
624	ps->callbacks = dm_vcalloc(ps->exceptions_per_area,
625				   sizeof(*ps->callbacks));
626	if (!ps->callbacks)
627		return -ENOMEM;
628
629	/*
630	 * Do we need to setup a new snapshot ?
631	 */
632	if (new_snapshot) {
633		r = write_header(ps);
634		if (r) {
635			DMWARN("write_header failed");
636			return r;
637		}
638
639		ps->current_area = 0;
640		zero_memory_area(ps);
641		r = zero_disk_area(ps, 0);
642		if (r)
643			DMWARN("zero_disk_area(0) failed");
644		return r;
645	}
646	/*
647	 * Sanity checks.
648	 */
649	if (ps->version != SNAPSHOT_DISK_VERSION) {
650		DMWARN("unable to handle snapshot disk version %d",
651		       ps->version);
652		return -EINVAL;
653	}
654
655	/*
656	 * Metadata are valid, but snapshot is invalidated
657	 */
658	if (!ps->valid)
659		return 1;
660
661	/*
662	 * Read the metadata.
663	 */
664	r = read_exceptions(ps, callback, callback_context);
665
666	return r;
667}
668
669static int persistent_prepare_exception(struct dm_exception_store *store,
670					struct dm_exception *e)
671{
672	struct pstore *ps = get_info(store);
673	sector_t size = get_dev_size(dm_snap_cow(store->snap)->bdev);
674
675	/* Is there enough room ? */
676	if (size < ((ps->next_free + 1) * store->chunk_size))
677		return -ENOSPC;
678
679	e->new_chunk = ps->next_free;
680
681	/*
682	 * Move onto the next free pending, making sure to take
683	 * into account the location of the metadata chunks.
684	 */
685	ps->next_free++;
686	skip_metadata(ps);
687
688	atomic_inc(&ps->pending_count);
689	return 0;
690}
691
692static void persistent_commit_exception(struct dm_exception_store *store,
693					struct dm_exception *e, int valid,
694					void (*callback) (void *, int success),
695					void *callback_context)
696{
697	unsigned int i;
698	struct pstore *ps = get_info(store);
699	struct core_exception ce;
700	struct commit_callback *cb;
701
702	if (!valid)
703		ps->valid = 0;
704
705	ce.old_chunk = e->old_chunk;
706	ce.new_chunk = e->new_chunk;
707	write_exception(ps, ps->current_committed++, &ce);
708
709	/*
710	 * Add the callback to the back of the array.  This code
711	 * is the only place where the callback array is
712	 * manipulated, and we know that it will never be called
713	 * multiple times concurrently.
714	 */
715	cb = ps->callbacks + ps->callback_count++;
716	cb->callback = callback;
717	cb->context = callback_context;
718
719	/*
720	 * If there are exceptions in flight and we have not yet
721	 * filled this metadata area there's nothing more to do.
722	 */
723	if (!atomic_dec_and_test(&ps->pending_count) &&
724	    (ps->current_committed != ps->exceptions_per_area))
725		return;
726
727	/*
728	 * If we completely filled the current area, then wipe the next one.
729	 */
730	if ((ps->current_committed == ps->exceptions_per_area) &&
731	    zero_disk_area(ps, ps->current_area + 1))
732		ps->valid = 0;
733
734	/*
735	 * Commit exceptions to disk.
736	 */
737	if (ps->valid && area_io(ps, REQ_OP_WRITE,
738				 REQ_PREFLUSH | REQ_FUA | REQ_SYNC))
739		ps->valid = 0;
740
741	/*
742	 * Advance to the next area if this one is full.
743	 */
744	if (ps->current_committed == ps->exceptions_per_area) {
745		ps->current_committed = 0;
746		ps->current_area++;
747		zero_memory_area(ps);
748	}
749
750	for (i = 0; i < ps->callback_count; i++) {
751		cb = ps->callbacks + i;
752		cb->callback(cb->context, ps->valid);
753	}
754
755	ps->callback_count = 0;
756}
757
758static int persistent_prepare_merge(struct dm_exception_store *store,
759				    chunk_t *last_old_chunk,
760				    chunk_t *last_new_chunk)
761{
762	struct pstore *ps = get_info(store);
763	struct core_exception ce;
764	int nr_consecutive;
765	int r;
766
767	/*
768	 * When current area is empty, move back to preceding area.
769	 */
770	if (!ps->current_committed) {
771		/*
772		 * Have we finished?
773		 */
774		if (!ps->current_area)
775			return 0;
776
777		ps->current_area--;
778		r = area_io(ps, REQ_OP_READ, 0);
779		if (r < 0)
780			return r;
781		ps->current_committed = ps->exceptions_per_area;
782	}
783
784	read_exception(ps, ps->area, ps->current_committed - 1, &ce);
785	*last_old_chunk = ce.old_chunk;
786	*last_new_chunk = ce.new_chunk;
787
788	/*
789	 * Find number of consecutive chunks within the current area,
790	 * working backwards.
791	 */
792	for (nr_consecutive = 1; nr_consecutive < ps->current_committed;
793	     nr_consecutive++) {
794		read_exception(ps, ps->area,
795			       ps->current_committed - 1 - nr_consecutive, &ce);
796		if (ce.old_chunk != *last_old_chunk - nr_consecutive ||
797		    ce.new_chunk != *last_new_chunk - nr_consecutive)
798			break;
799	}
800
801	return nr_consecutive;
802}
803
804static int persistent_commit_merge(struct dm_exception_store *store,
805				   int nr_merged)
806{
807	int r, i;
808	struct pstore *ps = get_info(store);
809
810	BUG_ON(nr_merged > ps->current_committed);
811
812	for (i = 0; i < nr_merged; i++)
813		clear_exception(ps, ps->current_committed - 1 - i);
814
815	r = area_io(ps, REQ_OP_WRITE, REQ_PREFLUSH | REQ_FUA);
816	if (r < 0)
817		return r;
818
819	ps->current_committed -= nr_merged;
820
821	/*
822	 * At this stage, only persistent_usage() uses ps->next_free, so
823	 * we make no attempt to keep ps->next_free strictly accurate
824	 * as exceptions may have been committed out-of-order originally.
825	 * Once a snapshot has become merging, we set it to the value it
826	 * would have held had all the exceptions been committed in order.
827	 *
828	 * ps->current_area does not get reduced by prepare_merge() until
829	 * after commit_merge() has removed the nr_merged previous exceptions.
830	 */
831	ps->next_free = area_location(ps, ps->current_area) +
832			ps->current_committed + 1;
833
834	return 0;
835}
836
837static void persistent_drop_snapshot(struct dm_exception_store *store)
838{
839	struct pstore *ps = get_info(store);
840
841	ps->valid = 0;
842	if (write_header(ps))
843		DMWARN("write header failed");
844}
845
846static int persistent_ctr(struct dm_exception_store *store, char *options)
847{
848	struct pstore *ps;
849	int r;
850
851	/* allocate the pstore */
852	ps = kzalloc(sizeof(*ps), GFP_KERNEL);
853	if (!ps)
854		return -ENOMEM;
855
856	ps->store = store;
857	ps->valid = 1;
858	ps->version = SNAPSHOT_DISK_VERSION;
859	ps->area = NULL;
860	ps->zero_area = NULL;
861	ps->header_area = NULL;
862	ps->next_free = NUM_SNAPSHOT_HDR_CHUNKS + 1; /* header and 1st area */
863	ps->current_committed = 0;
864
865	ps->callback_count = 0;
866	atomic_set(&ps->pending_count, 0);
867	ps->callbacks = NULL;
868
869	ps->metadata_wq = alloc_workqueue("ksnaphd", WQ_MEM_RECLAIM, 0);
870	if (!ps->metadata_wq) {
871		DMERR("couldn't start header metadata update thread");
872		r = -ENOMEM;
873		goto err_workqueue;
874	}
875
876	if (options) {
877		char overflow = toupper(options[0]);
878		if (overflow == 'O')
879			store->userspace_supports_overflow = true;
880		else {
881			DMERR("Unsupported persistent store option: %s", options);
882			r = -EINVAL;
883			goto err_options;
884		}
885	}
886
887	store->context = ps;
888
889	return 0;
890
891err_options:
892	destroy_workqueue(ps->metadata_wq);
893err_workqueue:
894	kfree(ps);
895
896	return r;
897}
898
899static unsigned persistent_status(struct dm_exception_store *store,
900				  status_type_t status, char *result,
901				  unsigned maxlen)
902{
903	unsigned sz = 0;
904
905	switch (status) {
906	case STATUSTYPE_INFO:
907		break;
908	case STATUSTYPE_TABLE:
909		DMEMIT(" %s %llu", store->userspace_supports_overflow ? "PO" : "P",
910		       (unsigned long long)store->chunk_size);
911	}
912
913	return sz;
914}
915
916static struct dm_exception_store_type _persistent_type = {
917	.name = "persistent",
918	.module = THIS_MODULE,
919	.ctr = persistent_ctr,
920	.dtr = persistent_dtr,
921	.read_metadata = persistent_read_metadata,
922	.prepare_exception = persistent_prepare_exception,
923	.commit_exception = persistent_commit_exception,
924	.prepare_merge = persistent_prepare_merge,
925	.commit_merge = persistent_commit_merge,
926	.drop_snapshot = persistent_drop_snapshot,
927	.usage = persistent_usage,
928	.status = persistent_status,
929};
930
931static struct dm_exception_store_type _persistent_compat_type = {
932	.name = "P",
933	.module = THIS_MODULE,
934	.ctr = persistent_ctr,
935	.dtr = persistent_dtr,
936	.read_metadata = persistent_read_metadata,
937	.prepare_exception = persistent_prepare_exception,
938	.commit_exception = persistent_commit_exception,
939	.prepare_merge = persistent_prepare_merge,
940	.commit_merge = persistent_commit_merge,
941	.drop_snapshot = persistent_drop_snapshot,
942	.usage = persistent_usage,
943	.status = persistent_status,
944};
945
946int dm_persistent_snapshot_init(void)
947{
948	int r;
949
950	r = dm_exception_store_type_register(&_persistent_type);
951	if (r) {
952		DMERR("Unable to register persistent exception store type");
953		return r;
954	}
955
956	r = dm_exception_store_type_register(&_persistent_compat_type);
957	if (r) {
958		DMERR("Unable to register old-style persistent exception "
959		      "store type");
960		dm_exception_store_type_unregister(&_persistent_type);
961		return r;
962	}
963
964	return r;
965}
966
967void dm_persistent_snapshot_exit(void)
968{
969	dm_exception_store_type_unregister(&_persistent_type);
970	dm_exception_store_type_unregister(&_persistent_compat_type);
971}
972